CSF images fast recognition model based on improved convolu-tional Neural Network
- DOI
- 10.2991/amcce-15.2015.97How to use a DOI?
- Keywords
- convolution neural network; cerebrospinal fluid; rectie function; identification; classification of linear support vector machine
- Abstract
The sparseness of feature is an important characteristic determining feature, which directly affects the accuracy of image recognition[1]. By studying the traditional con-volution neural network, we find that the learning of image features of cerebrospinal fluid cell easily overfits, but using rectie activation function instead of sigmoid activa-tion functions, the features extracted are more sparse and have faster convergence rate in the process of training. Then features extracted are classified through a linear sup-port vector machine. The experiments show that the improved model can enhance significantly the image recognition efficiency of cerebrospinal fluid, where two, three, four categories are respectively increased by 9.78%, 6.53%, 11.69%, and the average recognition time of a single image is also reduced 0.32s.
- Copyright
- © 2015, the Authors. Published by Atlantis Press.
- Open Access
- This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
Cite this article
TY - CONF AU - Wenming Huang AU - Jinqiang Leng AU - Zhenrong Deng PY - 2015/04 DA - 2015/04 TI - CSF images fast recognition model based on improved convolu-tional Neural Network BT - Proceedings of the 2015 International Conference on Automation, Mechanical Control and Computational Engineering PB - Atlantis Press SP - 518 EP - 524 SN - 1951-6851 UR - https://doi.org/10.2991/amcce-15.2015.97 DO - 10.2991/amcce-15.2015.97 ID - Huang2015/04 ER -